Background:  Disorders of glucose metabolism and their risk factors have been extensively studied in HIV-infected adults. Pediatric data on this topic are limited, controversial, and obtained mainly by cross-sectional studies.

Methods:  Insulin and glucose during 3-hour oral glucose tolerance test (OGTT) were measured at baseline and yearly for 4 years in 37 HIV-infected children receiving a protease inhibitor (PI) -based HAART containing lamivudine (3TC)/stavudine (d4T) + indinavir (IDV) or ritonavir (RTV) or nelfinavir (NFV) or a NNRTI-based HAART containing 3TC/tenofovir (TDF)/efavirenz (EFV). Age, weight, height, body mass index, Tanner pubertal stage, CD4%, viral load, and type and duration of HAART were recorded at baseline and yearly for 4 years. Generalized estimating equations were used to study the relationship between the log-transformed area under the curve of OGTT-insulin (I-AUC), time, type of HAART and potential confounders (gender, age at baseline, pubertal stage, z-score of body mass index, CD4%). The 3TC/d4T/IDV regimen was arbitrarily chosen as the HAART reference group and the effects of other regimens were modeled against it.

Results:  The patients were 4 to 16 years old at baseline. There were 37 until the second year, 35 at the third year, and 31 at the fourth year. During the study, HIV RNA was undetectable in most patients and the median value of CD4% increased. At baseline, PI-based regimens were administered to 35 of 37 patients (IDV n = 16; RTV n = 13; NFV n = 6), but their use decreased progressively at the third (2 of 33) and fourth (2 of 29) year of follow-up in favor of 3TC/TDF/EFV. The median values of I-AUC increased until the second year and decreased thereafter. The changes in glucose-AUC were more limited. None of the patients had diabetes mellitus or impaired glucose tolerance at any time of the study. Tanner stage ≥4 was strongly and positively associated with I-AUC (regression coefficient: 1.32; 95%CI 0.91 to 1.74; p <0.001). Among HAART regimens, NFV/3TC/d4T (regression coefficient –0.69; 95%CI –1.29 to –0.09; p <0.05) and EFV/3TC/TDF (regression coefficient –0.93; 95%CI -1.82 to -0.03; p< 0.05) but not RTV/3TC/d4T (regression coefficient:  0.29; 95%CI –0.85 to 0.26; p = 0.297) were associated with lower insulin levels as compared to IDV/3TC/d4T.

Conclusions:  Our 4-year follow-up study of HIV-infected children shows that HAART regimens containing NFV/3TC/d4T or EFV/3TC/TDF are associated with higher insulin sensitivity as compared with IDV/3TC/d4T and that puberty is a primary determinant of decreased insulin sensitivity.